Heat Diffusion Long-Short Term Memory Learning for 3D Shape Analysis

We aim to develop a 3D shape representation by utilizing the heat flows on 3D surfaces and the corresponding temporal dynamics of the heat flows within the diffusion period. We employ LSTM to capture the temporal dynamics of heat flows and extract joint information between different time-steps that are either consecutive or with a large interval. We incorporate a 3-layer fully-connected neural network with LSTM, and extend the proposed approach to a cross-domain scenario.